[ishaansathaye]

The experience

Adds Experiment → Teach a pathway: probe one of three cues, pair its arrival with a target spike, and probe again to discover a lasting change. With the default 10 ms cue-first protocol, cue A changes from a −57.0 mV peak with no spike to approximately −48.5 mV with a target spike. Reversing the order weakens the connection; cues not previously paired remain controls.

Four branching cells float in the existing dark scene. Traveling pulses come from the model's spike events and 10 ms transmission delay; contact size encodes efficacy. A compact dock exposes cue selection, order, gap, probe, pairing, pause, cancel and reset. The first probe supplies a comparison marker; older results are explicitly historical after retraining.

Evidence opens in the existing left-side tabbed reader. HDR/SDR glow, grayscale, reduced motion, keyboard controls, WebGL2 fallback, orbit and zoom are preserved. Short portrait and landscape layouts keep the controls accessible. State stays in memory for this visit, pauses when hidden or left, and resets on reload.

Review order

Stacked on #1: review the model/research PR first. This PR targets feat/timing-model so its diff contains the website integration. After owner approval of #1, retarget this PR to main and rerun checks before merging. Neither PR is authorized for automatic merge or deployment.

The research contract and next-phase roadmap are in docs/NEXT_PHASE.md; usage is in the existing guide. Phloem supplied an interaction principle, while this experiment's model, geometry and UI are original. The visual cells are schematic and the pedagogical model is uncalibrated; it does not predict personal learning, health or chemical concentrations.

Validation and review

• Full build, all 15 Rust tests, content validation (24 source records) and browser-script syntax pass.
• New end-to-end test uses the actual controls on WebGPU and WebGL2: baseline → pairing → successful probe, untouched cue, reversed pairing, exact pause-frame freeze, scene retention, cancellation/reset, source-tab return and identical Wasm batching.
• Mobile keyboard, reduced motion, grayscale, exclusive evidence/tools panels, 390×700 portrait and 844×390 landscape at DPR 1 and 2 are checked. Actual desktop/mobile screenshots were visually reviewed.
• Existing browser, refinement and panel suites pass. HDR checks pass with an actual float16 pipeline and emulated display-capability signal; physical luminance is not measured by this test.
• Portable static package passes under /doke/ with its intended CSP, including an experiment probe and fine anatomy loading.
• On this Mac at 1440×1000, active experiment frame intervals: 16.7 ms median, 17.7 ms p95 WebGPU / 17.5 ms p95 WebGL2, 180 sampled frames per backend. These are animation-frame measurements, not GPU timings or a guarantee for other devices.
• Parent reviewed the implementation and scientific wording; separate neuroscience and rendering agents checked model/visual agreement. Review caught and fixed stale probe labeling, a WebGPU shader numerical artifact, voltage normalization and compact-layout overlap.
• Candidate public files pass Gitleaks; generated assets, raw captures and local deployment settings remain ignored.
• GitHub model-and-build and GitGuardian checks pass for this PR. A real-time baseline → pairing → probe → evidence recording is retained locally with the ignored review captures.

Awaiting the project owner's visual and interaction review. The local preview opens directly at http://127.0.0.1:4175/?mode=experiment.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.